Xuan Li
**Describe the bug** If taichi is initialized with GPU, Torch cannot execute backward. PyTorch version: 1.12.0 **To Reproduce** ```py import taichi as ti import torch device = torch.device("cuda:0") ti.init(arch=ti.gpu) x...
To reproduce: ``` import warp as wp wp.init() @wp.kernel def test_grad(a: wp.array(dtype=wp.vec3), b: wp.array(dtype=wp.vec3)): tid = wp.tid() ai = a[tid] # not working tmp = wp.vec3(0., 0., 0.) tmp[0] +=...
### Bug Description In ```warp/sim/collide.py```, ```wp.clone``` are used to create new arrays when ```requires_grad=True```: ``` model.soft_contact_body_pos = wp.clone(model.soft_contact_body_pos) model.soft_contact_body_vel = wp.clone(model.soft_contact_body_vel) model.soft_contact_normal = wp.clone(model.soft_contact_normal) ... model.rigid_contact_point0 = wp.clone(model.rigid_contact_point0) model.rigid_contact_point1 =...